Hello,
When quantization is performed, the output values in the output layer become sparse values with Effective bits = 4. Is there a way to obtain denser output values like Effective bits = 8?
The number of outputs around 128 is small, but it is important for the application to obtain those values.